VesiclePy Segmentation Model Training and Performance Metrics

A 5.5 KB Excel dataset contains performance metrics for VesiclePy, an automated pipeline for segmenting and classifying neuronal vesicles from Volume Electron Microscopy data. The dataset, created by Jason Ken Adhinarta and last updated in May 2026, includes results from processing a multi-terabyte serial EM dataset of Hydra vulgaris, annotating 53,851 vesicles from 20 complete neurons.

Use Cases

• Benchmarking segmentation model performance based on ground truth manual annotations mentioned in the description
• Analyzing vesicle classification accuracy for 5 distinct vesicle types identified by the pipeline
• Evaluating the spatial analysis capabilities of VesiclePy based on unique 3D location data for each vesicle
• Clustering neurons into subtypes based on combined vesicle data and morphological information

Strengths

• Includes ground truth manual annotations for quantifying performance
• Covers 53,851 vesicles from 20 complete neurons
• Classifies vesicles into 5 distinct types

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment
• The dataset is very small (5.5 KB), suggesting it contains summary metrics rather than raw data

Provenance

Source

figshare

Collection Method

Generated by the VesiclePy pipeline using high-pressure frozen serial EM data of Hydra vulgaris.

Freshness

Last updated 2026-05-14 17:51:52; freshness should be verified